CCA is most appropriate when the proportion of missing data is relatively small and when the data are missing completely at random (MCAR). If these conditions are met, the bias introduced by excluding incomplete cases is minimized, and the loss of statistical power is not substantial.
